def _test_complete_flow(self, train_input_fn, eval_input_fn,
predict_input_fn, input_dimension, label_dimension,
batch_size):
feature_columns = [
feature_column.numeric_column('x', shape=(input_dimension, ))
]
est = dnn.DNNEstimatorV2(head=regression_head.RegressionHead(
label_dimension=label_dimension),
hidden_units=(2, 2),
feature_columns=feature_columns,
model_dir=self._model_dir)
# Train
num_steps = 10
est.train(train_input_fn, steps=num_steps)
# Evaluate
scores = est.evaluate(eval_input_fn)
self.assertEqual(num_steps, scores[ops.GraphKeys.GLOBAL_STEP])
self.assertIn('loss', six.iterkeys(scores))
# Predict
predictions = np.array([
x[prediction_keys.PredictionKeys.PREDICTIONS]
for x in est.predict(predict_input_fn)
])
self.assertAllEqual((batch_size, label_dimension), predictions.shape)
# Export
feature_spec = feature_column.make_parse_example_spec(feature_columns)
serving_input_receiver_fn = export.build_parsing_serving_input_receiver_fn(
feature_spec)
export_dir = est.export_saved_model(tempfile.mkdtemp(),
serving_input_receiver_fn)
self.assertTrue(gfile.Exists(export_dir))
def _dnn_estimator_fn(weight_column=None, label_dimension=1, **kwargs):
"""Returns a DNNEstimator that uses regression_head."""
return dnn.DNNEstimatorV2(
head=regression_head.RegressionHead(
weight_column=weight_column,
label_dimension=label_dimension,
# Tests in core (from which this test inherits) test the sum loss.
loss_reduction=losses_utils.ReductionV2.SUM_OVER_BATCH_SIZE),
**kwargs)
def _test_ckpt_converter(self, train_input_fn, eval_input_fn,
predict_input_fn, input_dimension,
label_dimension, batch_size, optimizer):
# Create checkpoint in CannedEstimator v1.
feature_columns_v1 = [
feature_column._numeric_column('x', shape=(input_dimension, ))
]
est_v1 = dnn.DNNEstimator(
head=head_lib._regression_head(label_dimension=label_dimension),
hidden_units=(2, 2),
feature_columns=feature_columns_v1,
model_dir=self._old_ckpt_dir,
optimizer=optimizer)
# Train
num_steps = 10
est_v1.train(train_input_fn, steps=num_steps)
self.assertIsNotNone(est_v1.latest_checkpoint())
self.assertTrue(est_v1.latest_checkpoint().startswith(
self._old_ckpt_dir))
# Convert checkpoint from v1 to v2.
source_checkpoint = os.path.join(self._old_ckpt_dir, 'model.ckpt-10')
source_graph = os.path.join(self._old_ckpt_dir, 'graph.pbtxt')
target_checkpoint = os.path.join(self._new_ckpt_dir, 'model.ckpt-10')
checkpoint_converter.convert_checkpoint('dnn', source_checkpoint,
source_graph,
target_checkpoint)
# Create CannedEstimator V2 and restore from the converted checkpoint.
feature_columns_v2 = [
tf.feature_column.numeric_column('x', shape=(input_dimension, ))
]
est_v2 = dnn.DNNEstimatorV2(head=regression_head.RegressionHead(
label_dimension=label_dimension),
hidden_units=(2, 2),
feature_columns=feature_columns_v2,
model_dir=self._new_ckpt_dir,
optimizer=optimizer)
# Train
extra_steps = 10
est_v2.train(train_input_fn, steps=extra_steps)
self.assertIsNotNone(est_v2.latest_checkpoint())
self.assertTrue(est_v2.latest_checkpoint().startswith(
self._new_ckpt_dir))
# Make sure estimator v2 restores from the converted checkpoint, and
# continues training extra steps.
self.assertEqual(
num_steps + extra_steps,
est_v2.get_variable_value(tf.compat.v1.GraphKeys.GLOBAL_STEP))
def _dnn_estimator_classifier_fn(n_classes=3, **kwargs):
return dnn.DNNEstimatorV2(
head=multi_class_head.MultiClassHead(
n_classes=n_classes),
**kwargs)
def _dnn_estimator_classifier_fn(n_classes=3, **kwargs):
return dnn.DNNEstimatorV2(
head=head_lib._multi_class_head_with_softmax_cross_entropy_loss(
n_classes=n_classes),
**kwargs)
